The best teams and people I have worked with are deeply curious.
They value curiosity, protect it and, most importantly, give it somewhere to go.
Curiosity without a safe place to play stays hypothetical - a dream, a hope, a whisper at a conference that get’s forgotten. People share interesting links, attend an AI demonstration or spend an energetic afternoon at a hackathon. Then Monday arrives, the prototype cannot be hosted, the model cannot be accessed and the idea quietly returns to a slide deck.
Curiosity is the instinct. A safe place to play turns it into capability.
The playground needs a fence
When I talk about a safe place to play, I mean infrastructure with thoughtful boundaries: enough freedom to explore and enough care to keep the space trustworthy. I banged on about this to no avail at several large law firms - some people got it. Others wanted them to play elsewhere. And this is the thing, they will and it’s better if you give them that place so they don’t inadvertently use shadow AI or other potentially harmful practices.
Core IT and Engineering teams have the leading role here. They can provide secure hosting, identity and access controls, approved data routes and a supported way to deploy small applications. They can give people secure access to capable LLMs—ideally frontier models where the use case and risk permit it—without asking every curious user to become a security architect first.
Those controls are the playground fence. The fence makes the space trustworthy. Inside it, people should have room to try things, make mistakes, show unfinished work and learn what the technology can actually do.
This gives meat on the bone to a hackathon. The play feels real because the prototype lives in an environment that resembles reality. It can use approved models, meet actual users and—if it proves useful—travel towards production without being rebuilt from nothing.
Most experiments will quite rightly end there. The promising ones should have a route forwards.
The dilemma is accelerating
Clayton Christensen’s The Innovator’s Dilemma describes an uncomfortable truth: strong organisations can listen carefully to their best customers, improve established products and allocate resources rationally—and still miss the next wave of innovation.
Early disruptive ideas often look inferior by the measures the established organisation already values. They serve smaller or emerging needs, so their first versions are easier to dismiss than to fund. This can happen while everybody involved is making perfectly sensible decisions; the organisation’s processes keep directing attention towards the current business.
AI accelerates this dilemma because the cost and time required to make a credible new thing are collapsing. More people can test an idea. Rough tools can improve quickly. A capability that looks marginal today may find its use case before a traditional planning cycle has finished describing it.
In a world where almost anything can be built, organisations need to become very good at learning which ideas deserve to survive.
Play creates the capability to choose
I see safe play as part of strategy. It gives an organisation practical experience to draw on while the conditions around it are changing. Here’s a (potentially nauseating) climb for you - just hold tight you might enjoy it:
From play grow skills. People stop talking about AI in the abstract and learn how to frame a task, structure context, evaluate an answer and recognise failure.
From skills grows expertise. Patterns become visible. Teams learn which problems reward automation, which need better process design and which should remain deliberately human.
From expertise grows alignment. Lawyers, operators, engineers, security teams and leaders gain a shared language because they have examined the same real artefacts.
And from alignment comes momentum. Decisions become easier. Useful experiments move faster. Governance becomes more specific because the conversation is now grounded in evidence.
Curiosity is difficult to mandate - some people have it and some don’t. I probanly have another article aorund “it’s not hard, it’s just new” whcih I tell my students most weeks. But back to curiosity: you can’t mandate it but you absolutely can make it feel very welcome when it shows up.
If you are a GC or Head of Legal Ops, go work with your IT / Engineering team to give people capable tools* and clear boundaires. Give their experiments a genuine route into the world you serve. Then let play do what it has always done: turn uncertainty into understanding, and understanding into something useful.
*(not Co-pilot or your own internal LLM wrapper but real frontier models in established harnesses).